| 摘要 |
<p><P>PROBLEM TO BE SOLVED: To provide a state prediction method for process that automatically determines the suitable number of proximity data vectors for each request point vector, and stabilizes prediction precision. <P>SOLUTION: A state prediction method for process in which a database is created in which input vectors and output vectors consisting of observation data representing operation states of processes are accumulated in pairs, a proximity data vector similar to a request point vector consisting of an input vector corresponding to an output vector at a desired point of time of prediction is acquired from the database, and a local model is structured from the proximity data vector to find an output vector at the desired point of time of prediction includes: generating a plurality of proximity data sets storing proximity data vectors while changing the number of proximity data; performing main component analysis of a plurality of proximity data sets to calculate Q statistics for the request point vector by the plurality of proximity data sets; and selecting a proximity data set corresponding to minimum Q statistics so as to structure the local model. <P>COPYRIGHT: (C)2013,JPO&INPIT</p> |
